CC-Top: Constrained Clustering for Dynamic Topic Discovery
About this Event
Research on multi-class text classification of short texts mainly focuses on supervised (transfer) learning approaches, requiring a finite set of pre-defined classes which is constant over time. This talk covers deep constrained clustering (CC) as an alternative to supervised learning approaches in a setting with a dynamically changing number of classes, a task we introduce as dynamic topic discovery (DTD). (Link to the paper)
Jann Goschenhofer is a final-year PhD student at the working group for the SLDS chair at LMU Munich in cooperation with the Fraunhofer ADA Lovelace Center. Currently, he focuses on Constrained Clustering and Positive Unlabeled Learning.
Dr. Matthias Aßenmacher is a postdoctoral researcher at the Chair of SLDS chair and the NFDI Consortium for Business, Economic and Related Data (BERD@NFDI). In 2021, he finished his PhD focusing on Natural Language Processing. He works on a diverse set of NLP Applications, including Active Learning, Bias and multi-modal use cases.